Microlearning has emerged as a prominent instructional strategy, delivering content in concise, focused segments that match with contemporary preferences for brief, accessible learning (Kossen and Ooi, 2021). This approach empowers learners by offering flexibility in when and what they learn, effectively supporting self-directed education (Judijanto, 2025). Its formats have diversified beyond text to include short videos, interactive simulations, podcasts, and gamified elements—all designed to reduce cognitive load and enhance engagement (Al-Nasheri and Alhalafawy, 2023; McNeill and Fitch, 2022). Moreover, microlearning facilitates just-in-time learning and has demonstrated effectiveness across varied contexts such as corporate training, healthcare education, and language learning (Zhang and West, 2020; Silva et al., 2025; Skalka et al., 2021). However, traditional implementations often rely on static, non-adaptive content, which limits personalization and fails to address individual learner needs (Richardson et al., 2023). Richardson et al. (2023) investigated microlearning’s flexibility and reported that it enables learners to access relevant information on demand by eliminating the need for extended formal sessions. Supporting this, Alias and Razak (2025) demonstrated multiple benefits of microlearning techniques, such as enhancing learning outcomes and improving knowledge acquisition. Silva et al. (2025) conducted a systematic review examining microlearning delivered in brief modules, revealing its effectiveness in enhancing knowledge retention and linking concepts to practical applications. Similarly, Judijanto (2025) performed a bibliometric analysis exploring microlearning’s role in a lifelong learning, concluding that it significantly supports self-directed education. Research has demonstrated microlearning’s effectiveness in various contexts including corporate training (Zhang and West, 2020), healthcare education (Silva et al., 2025), and language learning (Skalka et al., 2021).

The proliferation of mobile devices has further facilitated just-in-time learning opportunities, with studies showing improved knowledge retention when content is accessible via smartphones (Nikkhoo et al., 2023).

Despite these advancements, traditional microlearning approaches often rely on static, predetermined content modules that lack adaptability to individual learner needs. This one-size-fits-all limitation reduces the potential for truly personalized learning experiences (Monib et al., 2024).

Generative AI represents a paradigm shift in educational technology, moving beyond content delivery systems to become dynamic content creation partners. Unlike previous educational technologies that primarily organized or presented existing content, generative AI systems create original educational materials—including text explanations, assessment items, visual aids, and interactive simulations—based on pedagogical parameters and learner data (Díaz Redondo et al., 2021).

Generations of AI in education applications include content generation. Studies examine AI’s ability to create customized learning materials at scale (Mogavi et al., 2024). Assessment is another application of generative AI, with research exploring AI-generated formative assessments and automated feedback systems (Rüdian and Pinkwart, 2023).

While research on generative AI in education is expanding, most studies focus on content creation capabilities rather than examining how AI integration transforms specific pedagogical approaches or affects multidimensional learning outcomes.

Generative AI is revolutionizing education, driving innovation in areas such as text, image, audio, and video generation (Díaz Redondo et al., 2021). Microlearning plays a crucial role in developing individualized educational experiences. By evaluating learner data, generative AI produces customized content that aligns with individual skill levels and preferences, boosting engagement and retention. It also facilitates dynamic learning pathways that evolve with student progress, optimizing educational results (Mogavi et al., 2024). Additionally, generative AI enhances efficiency in content creation, allowing scalable and creative solutions for educators. Its potential to inspire new teaching methodologies enables more interactive and immersive learning experiences (Slivnaya et al., 2023). As technology evolves, generative AI is poised to redefine education, helping educators address diverse learning needs while fostering a more effective and engaging academic environment (Ruiz-rojas et al., 2023).

The intersection of microlearning, micro-content, and generative AI marks a significant shift in modern education. Microlearning’s concise, targeted format has proven effectiveness in knowledge delivery and construction (Ruiz-Rojas et al., 2023), yet manually tailoring micro-content to individual learners poses scalability challenges. Generative AI addresses this by using deep learning to analyze learner data, preferences, behaviors, and performance, which enables the automated production of personalized microlearning materials (Mahendra and Killis, 2025). This ensures content is precisely adapted to each student’s needs and learning styles (Lee, 2023). Furthermore, generative AI continuously improves through iterative learning, staying responsive to educational advancements. By integrating generative AI, microlearning becomes more scalable, customizable, and adaptive, ultimately enhancing student engagement and academic success (Kohnke, 2023; Cao et al., 2025; Díaz Redondo et al., 2021).

Learner engagement is a multifaceted construct encompassing behavioral, emotional, and cognitive aspects (Shi et al., 2023). As a critical factor in the learning process, engagement significantly influences students’ learning (Guan et al., 2023; McKee and Ntokos, 2022). Behavioral engagement refers to observable actions and task participation (Draxler et al., 2022; Guan et al., 2023; Wong et al., 2024), while emotional engagement represents internal states that are not directly observable (Fredricks et al., 2016; Guan et al., 2023; Shi et al., 2023; Zhang and West, 2020). Emotional engagement relates to students’ affective responses, which positively impact achievement (Shi et al., 2023; Yin et al., 2021; Wang and Degol, 2014). Cognitive engagement reflects deep cognitive strategies that shape knowledge processing, retention, and application, ultimately strengthening problem-solving skills (Shi et al., 2023). When applied to microlearning design, these engagement frameworks enable educational content to adapt to individual learning preferences while fostering active participation and sustained involvement (Busse et al., 2020; Sahin and Kırmızıgül, 2023). However, few studies have examined how AI-assisted microlearning influences engagement. Previous studies investigated the influence of microlearning on engagement. A study conducted by McKee and Ntokos (2022) examined how microlearning affects online student engagement. Results showed that microlearning helps students to stay entirely focused and engaged more in learning by allowing them to complete instructional tasks depending on their own schedule rather than on someone else’s schedule. In contrast, Moore et al. (2024) argued that microlearning enhances students’ engagement and motivation, these two factors that are crucial for sustaining lifelong learning behaviors.

From reviewing the literature, it is evident that few studies have investigated how AI-based microlearning affects engagement. Moreover, this study novelty lies in the crucial and new topic it explores. Also, this study is unique compared to other studies as it utilizes experimental and qualitative design compared to other previous studies. Despite the growing body of research about microlearning, there remains a lack of clarity regarding the current state of the literature related to influence of AI-based microlearning on engagement in online environment.

The impact of using AI-assisted microlearning is investigated by a mixed method design; by both quantitative and qualitative approaches. An explanatory sequential design (Creswell and John, 2018) was employed. The design was chosen to quantify the impact of AI-assisted microlearning on student engagement, and to explore the lived experiences and contextual factors influencing engagement through qualitative insights.

This study utilized a quasi-experimental design and a phenomenology approach, where participants went through the experience of using AI-assisted microlearning and expressed their thoughts and perspectives. Both groups completed the pre-test of engagement scale. During 10 weeks of multimedia course, both groups followed the same sequence of participating in attending online class activities. During the period, students in both groups followed the sequence three times a week, resulting in a total of 42 online classes 50-min sessions. Several topics were taught, including multimedia components, multimedia significance, designs of using multimedia in teaching technology curriculum, and theories of design. All students completed the pre-test and post-test. The experimental group additionally participated in discussion forums and semi-structured interviews 1 week after the last session, as shown in Figure 1.

Students at the third-year and fourth-year study levels at Palestine Technical University–Kadoorie (PTUK) in a multimedia course from technology education department formed the sample for this study. They were enrolled in Multimedia in Educational Technology at PTUK during the fall semester of the 2025/2026 academic year. Selection criteria included: enrollment in the designated course section, completion of prerequisite courses in computer literacy and instructional design, no prior formal training in AI-assisted learning tools or microlearning platforms, and willingness to participate in online forums and optional interviews for qualitative data. Students were pre-service teachers preparing to teach technology curricula in schools. Technology education students who enrolled in this program itself aims to gain a deep understanding of current teaching strategies and theories to teach technology curriculum in schools. Technology education study program aims to strengthen technical skills, as well as to enhance designing, and production of educational technology products. Quasi experiment consisted of 50 students who were purposively selected from the same course taught by one instructor to minimize instructor-style effects. Purposive sampling was applied since a quasi-experiment was conducted (Salhab, 2024). Table 1 shows the demographic data of the respondents.

A power analysis was conducted using G*Power 3.1 (Faul et al., 2009) for a mixed-design ANCOVA with two groups, one covariate (pre-test score), and four engagement dimensions. With an assumed medium effect size (f = 0.25), α = 0.05, and power = 0.80, the recommended sample size was 44. A sample of 50 was selected to account for potential attrition and to ensure robust qualitative data from at least 20 interview participants (experimental group).

Since the sample was drawn from a single university, discipline (technology education), and cultural context (Palestine), findings may be limited in generalizability of findings to other populations, disciplines, or regions. Future multi-site and cross-disciplinary replications are recommended.

To mitigate internal validity threats, stratified random assignment was conducted based on academic year and prior GPA, and pre-test equivalence was checked via t-tests. Equivalency in pre-test scores was confirmed, as shown in Table 2. Another threat, which included history and maturation, was mitigated by making control group exposed to same timeline and external events and ANCOVA controlled for pre-test scores.

Two groups of students (N = 50) were assigned to one of two conditions, the experimental group (n = 25), who received AI-assisted microlearning modules, and a control group (n = 25) who received traditional online learning materials without AI personalization. A stratified random assignment was used within the purposively sampled cohort. Students were first stratified by academic year (third vs. fourth year) and prior GPA (above/below cohort median) to ensure balanced groups. Within each group, students were randomly assigned to either the experimental or control group using a random number generator. Participants were not informed of their group designation to reduce expectancy effects. They were told the study was investigating modern digital learning tools.

The same instructor taught both groups to control for teaching style, communication, and content delivery differences. The instructor’s role was standardized by providing identical weekly topics, objectives, and assignment deadlines to both groups. The instructor also participated in discussion forums with a standardized weekly presence and used a scripted set of discussion prompts to ensure consistency. To minimize bias, the instructor was not involved in data collection, analysis, or interpretation. A teaching assistant managed the AI tool integration and microlearning delivery for the experimental group to prevent differential treatment.

For discussion forums and semi structured interviews, 12 participants from third year students from the College of Arts and Social Sciences in Technology Education department at PTUK who were in the experimental group participated in the study. The selection criteria were based on knowledge and experience of participants who fluently uses online platforms and participants who passed the basic course of computer courses. Purposive sampling was used to select participants from the same course and instructor to control for instructor effect. Participants were randomly assigned to experimental or control groups.

This study was conducted for a multimedia course at PTUK in northern Palestine. It had six goals, including: identifying multimedia concept and components, designing an educational poster and a brochure about a lesson in technology school curriculum, creating educational videos about lessons in technology curriculum, and utilizing AI tools to design images and edit images. In addition, the course incorporated short content-videos, discussion, forum, and assignments. The AI tools used by faculty are listed in Table 3.

The study followed an explanatory sequential design, with data collection conducted in three sequential stages, as outlined by Bowen (2017). This approach involves first gathering and analyzing quantitative data, followed by qualitative data collection discussion forums and semi-structured interviews. This design aims to quantifying the phenomenon under investigation and then explore the findings in greater depth through qualitative insights.

As part of the quasi-experimental design, 50 students completed a pre-course engagement scale before the intervention (AI-assisted microlearning). Both the control and experimental groups completed the survey online, which offered a convenient, time-efficient, and cost-effective method for data collection (Herrmann-Abell et al., 2016). The pre-existing scale, originally designed to assess engagement toward Information and Communication Technology (ICT) (Creswell and Clarck, 2011), was adapted for micro learning. The scale comprised 20 items across three constructs: the cognitive dimension consisted of 7 items, the affective dimension was composed of 8 items, and the behavioral dimension consisted of 5 items. All items were measured on a 5-point Likert scale (1 = completely disagree, 5 = entirely agree).

Since the scale was translated from English to Arabic, its validity was assessed through expert review (face validity) by professors specializing in educational psychology, educational technology, and curriculum design. Pearson’s correlation was used to verify construct validity, with coefficients ranging between 0.621–0.882 for affective, 0.631–0.811 for behavioral, and 0.677–0.834 for cognitive dimensions—all statistically significant (p < 0.01), confirming strong validity (Rebouças et al., 2018).

Reliability was tested using Cronbach’s alpha, yielding high internal consistency: 0.868 (affective), 0.754 (behavioral), and 0.736 (cognitive), with an overall scale reliability of 0.834, indicating robust consistency (Oktavia et al., 2018).

Qualitative data were collected through semi-structured interviews. After AI-assisted microlearning intervention, 25 interviews were conducted with participants who voluntarily decided to participate. These participants met two criteria: (a) third-year college level or higher, and (b) enrollment in the multimedia course in 2025/2026. Sessions lasted 30–45 min, conducted via Zoom and in-person, with participant consent for recording. Data were also collected from 25 students’ forum comments in a multimedia course at PTUK, anonymized to protect identities. Semi-structured interviews included the following questions:

To uphold the study’s reliability and credibility, four key validation criteria were applied, including cross-referencing and expert validation. Forum responses were cross-checked with input from field experts during coding, resulting in an 85% inter-coder agreement rate, exceeding the standard 80% threshold. Credibility was ensured by verifying through code cross-checking, data verification, and discrepancy checks to ensure accuracy. Dependability was established through code–recode strategy. The dataset was independently coded twice with a 2-week gap between sessions to assess consistency. A stepwise replication approach was used, where both researchers analyzed data separately and compared outcomes. Transferability and conformability were checked by selecting purposive sampling that ensured participant selection aligned with study criteria. Researcher bias was minimized by strictly basing conclusions on participant narratives and maintaining detailed procedural records. A dual-check process was implemented throughout the study to confirm data accuracy.

The study received ethical approval from PTUK. Moreover, informed consent was obtained from all participants, confidentiality was maintained by anonymizing identities and storing data securely on a password-protected computer.

All students responded to the instructor’s questions and received comments from their peers on questions the instructor posted in the discussion forums, demonstrating learner–learner interaction. Students in this AI-assisted microlearning discussed and interacted more with peers. Comments illustrated this interaction, such as” I do agree with you,”” Yes, the video about image editing is really helpful…, “I do think that you are right when you mentioned google sites. “. Another participant added “the short videos made it easy to keep up. I’d watch one on the bus and actually do the exercises because they did not feel like a chore.” Also, another “The badges were silly but motivating—I wanted to ‘collect’ all the modules!.” “I commented more in forums because the topics were specific and short.” Also, there was another interaction with the microlearning videos. Examples of this interaction is: “This video explained lots of details in a short time and step by step guide to make a brochure from scratch,” “the posted video is very comprehensive and easy to follow,” “very helpful video, especially for multimedia course.” These findings indicate that AI-assisted microlearning was effective in stimulating different types of interaction, particularly peer interaction and content interaction.

AI-assisted microlearning significantly influenced emotional engagement by leveraging personalized, interactive, and adaptive learning techniques.

Social engagement is defined as a process in which learners connect with different technologies through learner–interface interaction to support further learning. In this study, students demonstrated interaction by connecting with different technologies, information, and peers.

Some students reported that the AI-based mini videos were interesting with high quality and resolution. They made the concepts easier and explained them in detail, as Ali mentioned “The sound is clear, the videos explain the information step by step, and explains in an easy and simple way, I can share information with my peers easily.” Also Shadi said “the videos taught me how to make education more effective, collaborative and meaningful.” This implies that the videos were effective and support information generation and content sharing.

Cognitive engagement refers to a process of connecting specialized nodes or information sources together. In this type of engagement, learners interact with the content and peers, exchanging video content and sometime describing or explaining it.

Students in interviews mentioned that AI-based microcontent helped them stay attentive and focused on tasks, since tasks were bite-sized and did not require long attention. Most students confirmed that AI-assisted microlearning videos and tasks captured their attention easily. Many students became more attentive to AI-assisted microlearning activities because they included a variety of activities and mini-interactive instructional videos.

Nadia mentioned, “AI-based micro-games drives me to stay focused and attentive.”

Moreover, integrating different micro-activities students to share knowledge and construct in peer interactions and pay attention to each other’s’ comments. Majd said “I would have more chances to share my thoughts instantly since tasks are short and vides are short, this allows me to save more attention from my side and spend more time ask questions and receive answers from my online classmates, reading their answers helped to think critically.”

This study is limited by its single-institution sample and focus on technology students, which may affect generalizability. The quasi-experimental design, while practical, lacks random assignment. Future replications should include randomized control trials across disciplines. Also, the study employed a relatively small sample size (N = 50) drawn from a single academic program (Technology Education) at one university. This limited sample reduces statistical power for detecting smaller effects and constrains the generalizability of findings to broader student populations. The homogeneity of participants—all pre-service technology teachers with similar digital competencies—further limits applicability to students from other disciplines, academic levels, or with varying technological proficiencies. Moreover, despite efforts to ensure group equivalence through stratified assignment, the study utilized a quasi-experimental design rather than true randomization. Participants were not randomly assigned from a larger population but were instead enrolled in a specific course section, introducing potential selection bias. Although statistical tests indicated no significant pre-intervention differences between groups, unmeasured confounding variables (e.g., intrinsic motivation, prior online learning experience) may have influenced outcomes.

Future studies should recruit participants from both technical and non-technical programs across 3–5 universities. While this study focused on multimedia students, replication with larger, randomized samples is recommended. Moreover, future research could extend the limits beyond Engagement and examine effects on deeper learning outcomes including critical thinking, problem-solving transfer, and metacognitive skill development. Also, equity considerations could be investigated by conducting studies that explore whether AI-assisted microlearning reduces or exacerbates achievement gaps for underrepresented or academically at-risk student populations.